IoT-Empowered Smart Agriculture: A Real-Time Light-Weight Embedded Segmentation System

Internet of Things (IoT) is an emerging technology where standalone equipments and autonomous devices are connected to each other and users via Internet. When IoT concept meets agriculture, the future of farming is pushed to the next level, giving birth to what is called "Smart Agriculture" or "Precision Agriculture". The most important benefit from IoT is that a user can daily monitor his crop online in a seamless fashion. High quality data gathered from various sensors and transferred wirelessly to farm database will increase farmers understanding to their landuse leading to increasing income and product quality. One of the monitoring process is weeds detection and crop yield estimation using camera sensors. The acquired images help farmers to build map of weeds distribution or yield quantity all over the field, these maps can be used either for real-time processing or to predetermine weeds regions based on field maps history of the previous seasons. This process is referred to as segmentation problem. Several algorithms have been proposed for that purpose, however, these algorithms were run only on high performance computers. In this paper, we evaluate performance and the robustness of the most used legacy algorithms under local conditions. We focused on implementing these schemes within real-time application constraint. For instance, these algorithms were implemented and run in a low-cost embedded system.